Research on the Application of Generative AI in Immersive Tasks for RPG Games

In recent years, generative artificial intelligence has developed rapidly, supported by improvements in computing resources, data availability, and model architecture design. The introduction of the Transformer architecture in 2017 provided a key technical basis for later large-scale language models. The release of ChatGPT in 2022 further increased the visibility and adoption of generative AI systems, contributing to their wider application beyond research settings. At present, research in this area has increasingly involved multimodal models and more efficient deployment strategies. In the field of role-playing games (RPGs), many existing task systems are still based on pre-designed scripts, with relatively fixed task structures, limited narrative variation, and constrained interaction patterns between non-player characters (NPCs). These characteristics limit the flexibility of gameplay experience to some extent and make it difficult to fully support diverse player interactions. This paper reviews studies on generative AI applications in RPGs since 2020, focusing on three aspects: narrative generation, dynamic task design, and NPC dialogue systems. It further summarizes current research developments, discusses issues such as content controllability, experience consistency, and computational cost, and outlines several research directions, including multimodal task modeling and personalization mechanisms.